CN112115830B - Target distributed fusion recognition method based on bit domain feature extraction - Google Patents
Target distributed fusion recognition method based on bit domain feature extraction Download PDFInfo
- Publication number
- CN112115830B CN112115830B CN202010945655.1A CN202010945655A CN112115830B CN 112115830 B CN112115830 B CN 112115830B CN 202010945655 A CN202010945655 A CN 202010945655A CN 112115830 B CN112115830 B CN 112115830B
- Authority
- CN
- China
- Prior art keywords
- neural network
- edge
- distributed
- target
- bit domain
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a target distributed fusion recognition method based on bit domain feature extraction, which is used in edge equipment and an edge domain fusion center, and comprises the following steps: 1) Obtaining sensing signals from various edge devices; 2) The method comprises the steps of obtaining a pre-constructed and trained deep neural network, splitting the deep neural network according to a preset deployment rule, and deploying the deep neural network into various edge equipment and an edge domain fusion center to obtain a distributed neural network architecture; the deep neural network deployed in the edge device generates bit domain feature vectors by extracting the bit domain features of the output signals, so as to perform data transmission 3) loading the sensing signals and obtain target recognition results. Compared with the prior art, the method and the device can preprocess the target signal on the edge equipment, obtain the quantized and compressed bit domain characteristics, perform data transmission, relieve the calculated storage pressure and the network bandwidth pressure of the cloud center, and improve the response speed of the target recognition process.
Description
Technical Field
The invention relates to the field of target identification, in particular to a target distributed fusion identification method based on bit domain feature extraction.
Background
The edge calculation refers to a novel calculation model capable of executing calculation at the edge of a network, the edge refers to any calculation and network model between paths from a data source to a cloud calculation center, and downlink data of the edge represents cloud service and uplink data represents everything interconnection service. In addition, the actual deployment of the edge computing naturally has the distributed characteristic, supports the distributed computing and storage, and has the capabilities of dynamic scheduling and unified management of distributed resources, supporting distributed intelligence, distributed security and the like. The edge computing can meet a plurality of key requirements of industry digitization in aspects of agility connection, real-time service, data optimization, application intelligence, security, privacy protection and the like. Meanwhile, the long-term development of the computing and storage technology enables the edge perception device to have strong computing capacity and can bear certain computing tasks, so that the response speed of the system is increased.
Artificial intelligence techniques, particularly deep learning methods, have been widely used in the field of image/video object detection and recognition, with good expectations. At present, some researches on target recognition by applying a neural network exist, and Yang Weichao uses a deep learning method to modulate and recognize signals in literature 'Alpha stable distribution noise communication signal modulation recognition research', so that a good recognition effect is obtained. Li Jia in the literature "study of digital modulation signal recognition method based on deep learning" proposes study of digital modulation signal recognition method based on deep learning, respectively applying two different deep learning models, and achieving the best effect by simulating the parameters required in the selected algorithm. None of the above studies considered the application of neural networks on the edge side for target recognition.
Disclosure of Invention
The invention aims to overcome the defect that the prior art does not consider that a neural network is applied to the edge side for target identification, and provides a target distributed fusion identification method based on bit domain feature extraction.
The aim of the invention can be achieved by the following technical scheme:
the target distributed fusion identification method based on bit domain feature extraction is used in an edge side communication network topology structure, wherein the edge side communication network topology structure comprises a plurality of edge devices and an edge domain fusion center, and the method comprises the following steps:
a signal acquisition step: obtaining a perception signal from the plurality of edge devices;
a neural network deployment step: the method comprises the steps of obtaining a pre-constructed and trained deep neural network, splitting the deep neural network according to a preset deployment rule, and deploying the deep neural network into the various edge devices and the edge domain fusion center to obtain a distributed neural network architecture;
a signal compression unit is arranged in the deep neural network deployed in the edge equipment, and the signal compression unit is used for extracting bit domain features of an output signal of the deep neural network deployed in the edge equipment to generate bit domain feature vectors;
the edge domain fusion center carries out target recognition according to the bit domain feature vector;
a target identification step: and loading the sensing signals into the distributed neural network architecture to obtain a target recognition result.
Further, the signal compression unit comprises a first signal compression layer and a second signal compression layer;
the first signal compression layer adopts a first activation function to map real values of each dimension of an output signal of the deep neural network deployed in the edge equipment to between 0 and 1;
the second signal compression layer is connected with the first signal compression layer, and maps real values of each dimension between 0 and 1 of the first signal compression layer to 0 and 1 by adopting a second activation function to generate a bit domain feature vector.
Further, the first activation function is a sigmoid activation function.
Further, the expression of the second activation function is:
where y is the output of the second activation function, x is the input of the second activation function, and ε is the adjustment threshold.
Further, in the neural network deployment step, the deployment rule is constructed according to the computing capability and the communication capability of the edge device.
Further, the method further comprises:
calculating average performance: repeatedly executing the target identification step until reaching a preset first time, and acquiring the average performance of the distributed neural network architecture;
and a neural network adjustment step: and adjusting the layer number and parameters of the distributed neural network architecture, repeatedly executing the signal acquisition step, the neural network deployment step, the target identification step and the average performance calculation step until reaching a preset second time, taking the distributed neural network architecture with the optimal average performance as an optimal distributed neural network architecture, and adopting the optimal distributed neural network architecture to perform target identification.
Further, the first number of times has a value in the range of 3 to 20.
Further, the second number of times has a value in the range of 50 to 150.
Further, the calculation index of average performance includes recognition speed and power consumption.
Compared with the prior art, the invention has the following advantages:
(1) The invention provides a distributed deep neural network architecture, which can preprocess a target signal on edge equipment and obtain quantized compressed bit domain characteristics, and meanwhile, the edge equipment only transmits compressed characteristic data to an edge domain fusion center;
the target identification calculation is carried out on the edge side, so that the calculation storage pressure of the cloud center can be relieved, and the response speed of the target identification process is improved;
by carrying out characteristic quantization compression on the target signal, bit domain characteristic data with smaller information quantity is obtained, data transmission is carried out, and the data quantity of network transmission is reduced, so that the transmission traffic of the edge equipment and the edge domain fusion center is effectively reduced, and the network bandwidth pressure is relieved.
(2) According to the invention, on the basis of considering the computing and communication capacities of the edge equipment and the edge domain fusion center, the optimal division of the neural network hidden layer of the distributed neural network architecture is realized by computing the average performance of the distributed neural network architecture after each division, so that the overall energy consumption and the processing time are reduced.
Drawings
FIG. 1 is a schematic overall flow chart of a target distributed fusion recognition method based on bit domain feature extraction;
FIG. 2 is a schematic diagram of data transmission of a target distributed fusion recognition method based on bit domain feature extraction according to the present invention;
FIG. 3 is a schematic diagram of a distributed deep neural network architecture according to the present invention;
FIG. 4 is a graph of test set accuracy for each round of iterations of the underlying neural network without using bit fields;
FIG. 5 is a graph of test set accuracy for each iteration of the improved neural network of the present invention based on bit domain feature extraction;
in the figure, epochs is the iteration number, accuracy is the accuracy, baseline is the base neural network curve, and trasformed is the modified neural network curve.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The present embodiment is implemented on the premise of the technical scheme of the present invention, and a detailed implementation manner and a specific operation process are given, but the protection scope of the present invention is not limited to the following examples.
Example 1
The embodiment provides a target distributed fusion recognition method based on bit domain feature extraction, which is used in an edge side communication network topology structure comprising various edge devices and an edge domain fusion center,
the method of this example is described below in terms of method overview, detailed implementation, specific applications, and result comparison.
1. Summary of the method
As shown in fig. 1, the target distributed fusion recognition method based on bit domain feature extraction in this embodiment includes the following steps:
signal acquisition step S1: obtaining sensing signals from various edge devices;
neural network deployment step S2: the method comprises the steps of obtaining a pre-constructed and trained deep neural network, splitting the deep neural network according to a preset deployment rule, and deploying the deep neural network into various edge equipment and an edge domain fusion center to obtain a distributed neural network architecture;
the deep neural network deployed in the edge equipment is provided with a signal compression unit, and the signal compression unit is used for extracting bit domain features of an output signal of the deep neural network deployed in the edge equipment to generate bit domain feature vectors;
and the edge domain fusion center performs target recognition according to the bit domain feature vector.
Target recognition step S3: and loading the sensing signals into a distributed neural network architecture to obtain a target identification result.
The steps are described in detail below.
1. Signal compression unit
The signal compression unit comprises a first signal compression layer and a second signal compression layer;
the first signal compression layer adopts a first activation function to map real values of each dimension of output signals of the deep neural network deployed in the edge equipment to between 0 and 1;
the second signal compression layer is connected with the first signal compression layer, and maps real values of each dimension between 0 and 1 of the first signal compression layer onto 0 and 1 by adopting a second activation function to generate bit domain feature vectors.
The first activation function is a sigmoid activation function.
The expression of the second activation function is:
where y is the output of the second activation function, x is the input of the second activation function, and ε is the adjustment threshold.
2. Deployment rules
In the neural network deployment step, deployment rules are constructed according to the computing capability and the communication capability of the edge equipment.
3. Other steps
The target distributed fusion identification method of the embodiment further comprises the following steps:
average performance calculation step S4: repeatedly executing the target identification step until reaching the preset first times, and acquiring the average performance of the distributed neural network architecture;
the neural network adjustment step S5: and adjusting the layer number and parameters of the distributed neural network architecture, repeating the steps of signal acquisition, neural network deployment, target identification and average performance calculation until the preset second times are reached, taking the distributed neural network architecture with the optimal average performance as the optimal distributed neural network architecture, and adopting the optimal distributed neural network architecture to perform target identification.
The first number of values is within the range of 3 to 20. The value range of the second times is within the range of 50 to 150. The calculation index of average performance includes recognition speed and power consumption.
2. Detailed implementation procedure
The target distributed fusion recognition method based on bit domain feature extraction provided by the embodiment has the following two features:
a. a distributed deep neural network architecture is presented that can preprocess a target signal at an edge device and obtain quantized compressed bit domain features. Meanwhile, the edge equipment only transmits the compressed characteristic data to the edge domain fusion center, so that the transmission traffic of the edge equipment and the edge domain fusion center is effectively reduced;
b. the neural network architecture optimally divides the hidden layers of the neural network on the basis of considering the computing and communication capacities of the edge equipment and the edge domain fusion center, so that the overall energy consumption and the processing time are reduced.
The method specifically comprises the following steps:
step 1: according to the topology structure of the edge side communication network, a distributed execution overall deep neural network model is designed. The part of the network model corresponding to the edge device needs to be able to quantize the compressed sensing signal, in order to achieve this, the penultimate layer of the edge device network uses a sigmoid activation function to map the real values of the input vector of the layer between (0, 1), and the last layer uses the following activation function to map the real values of the input vector of the layer onto 0 or 1:
wherein epsilon is a threshold value, and the specific value is required to be adjusted according to the actual application scene. The network at the edge domain fusion center can receive and fuse inputs from a plurality of edge device neural networks, and then output a final recognition result.
The floating point number is stored in the computer with the length of 4 bytes and 32 bits, and 0 and 1 are only one bit; mapping real values of each dimension onto 0 or 1 is equivalent to sacrificing accuracy in exchange for memory length. Even only 0 and 1 may represent information such as 0101101. But the test results also show that the recognition accuracy is not significantly degraded in practice.
Step 2: and training the model by using the marked data sample to ensure that the model has high recognition accuracy. And then splitting the model according to an edge side communication network topology structure, and deploying the model to each edge device and an edge domain fusion center.
Step 3: actual target recognition inferences are made. The specific flow is that different high-dimensional sensing signals obtained by each edge device are used as the input of a self-deployment neural network, and each neural network outputs a bit domain feature vector after being processed. And then, feature data on the edge equipment is transmitted to an edge domain fusion center through a communication network, the edge domain fusion center takes the multi-source heterogeneous features as the input of a self-deployed neural network, and an identification result is finally obtained after fusion processing.
Step 4: and repeatedly executing the step 4 for m times to obtain the average performance (recognition speed and power consumption) of the current overall depth neural network model, wherein m is generally [3,20] according to a statistical theory. And (3) adjusting the layer number and parameters of the neural network by taking the average performance as a reference, repeating the steps 1-4 (t E [50,150 ]) t times, and finally obtaining the overall depth neural network model with the optimal performance.
Adjusting parameters of the neural network includes dividing points between sub-networks and training parameters of the neural network.
3. Specific application
As shown in fig. 2, the target problem is defined first, and the problem solved by the invention is to effectively reduce the data transmission traffic of the edge device and the edge domain fusion center in the target object identification process by using the neural network based on the bit domain feature extraction. The target object refers to an objective object perceived by edge equipment, such as a fighter plane, a submarine and the like. Edge side refers to the network edge geographically relative to the cloud center. The edge side is composed of a plurality of edge devices, and the edge devices can detect and sense a target object to obtain electromagnetic signals (such as communication signals, radar signals, photoelectric signals and the like). In the information acquisition process, the edge device first obtains a structured scout data vector x= (x) p ,x m ) Wherein the subvector x p Representing the relevant information of the acquisition node (edge equipment), including node type, position, sampling time, etc., sub-vector x m Characteristic parameter information representing the target object, the characteristic parameters of which are different for different types of electromagnetic signals. For example, the characteristic parameters of the communication signal mainly include: operating frequency band, carrier frequency, modulation pattern, signal duration, power level, communication regime, transmitter position, etc.; the radar signal characteristic parameters mainly comprise: radio frequency and its variation characteristics, pulse width and its modulation characteristics, intra-pulse frequency or phase modulation characteristics, antenna scan type, scan period, pattern and polarization characteristics, etc.; the characteristic parameters of the photoelectric signals are mainly tensor data in the form of images or videos, etc. Different electromagnetic fieldsThe signals represent target objects with different properties, so that the target objects can be deduced back by analyzing the electromagnetic signals.
The edge device is often responsible for acquiring the electromagnetic signal and transmitting the electromagnetic signal to the edge domain fusion center, however, the process has the defects of high transmission pressure and high response delay. The scheme of the invention can carry out quantization compression on the data output by the edge equipment to obtain the bit domain characteristic data with smaller information quantity, and then the characteristic data is transmitted to the edge domain, thereby achieving the purposes of relieving bandwidth pressure and improving the real-time performance of the system.
The following is a specific implementation step of a target distributed fusion recognition method based on bit domain feature extraction:
step 1: according to the topology structure of the edge side communication network, a distributed execution overall deep neural network model is designed. The part of the network model corresponding to the edge device needs to be able to quantize the compressed sensing signal, in order to achieve this, the penultimate layer of the edge device network uses a sigmoid activation function to map the real values of the input vector of the layer between (0, 1), and the last layer uses the following activation function to map the real values of the input vector of the layer onto 0 or 1:
wherein epsilon is a threshold value, and the specific value is required to be adjusted according to the actual application scene. The network at the edge domain fusion center can receive and fuse inputs from a plurality of edge device neural networks, and then output a final recognition result.
Step 2: and training the model by using the marked data sample to ensure that the model has high recognition accuracy. And then splitting the model according to an edge side communication network topology structure, and deploying the model to each edge device and an edge domain fusion center.
Step 3: actual target recognition inferences are made. The specific flow is that different high-dimensional sensing signals obtained by each edge device are used as the input of a self-deployment neural network, and each neural network outputs a bit domain feature vector after being processed. And then, feature data on the edge equipment is transmitted to an edge domain fusion center through a communication network, the edge domain fusion center takes the multi-source heterogeneous features as the input of a self-deployed neural network, and an identification result is finally obtained after fusion processing.
Step 4: and repeatedly executing the step 4 for m times to obtain the average performance (recognition speed and power consumption) of the current overall depth neural network model, wherein m is generally [3,20] according to a statistical theory. And (3) adjusting the layer number and parameters of the neural network by taking the average performance as a reference, repeating the steps 1-4 (t E [50,150 ]) t times, and finally obtaining the overall depth neural network model with the optimal performance.
4. Comparison of results
Fig. 2 depicts a workflow of the distributed neural network target fusion recognition based on edge computation, and fig. 3 presents a network structure schematic diagram of the distributed neural network. Based on fig. 2 and 3, the present invention implements an example of the above method and proposes an improvement scheme, while comparing the final recognition result with the basic neural network target fusion recognition method that does not use the bit field.
Let the number of data bits occupied by a real number be B; the dimension of the edge device output vector is m.
Therefore, the data transmission quantity of the edge equipment of the basic neural network and the edge domain fusion center when the bit domain is not used is as follows: m.B; the data transmission quantity of the improved neural network edge equipment and the edge domain fusion center based on the bit domain feature extraction in the scheme is as follows: m.1=m. The two are compared to obtain
If b=32, there is +.>
I.e. in this case the present solution reduces the data transmission traffic to the original model data3.13% of the traffic is transmitted.
Fig. 4 and 5 compare the prediction accuracy of the distributed neural network based on edge computation with that of the base neural network without using the bit domain, the left graph shows the test set accuracy of the base neural network without using the bit domain in each round of iteration, and the right graph shows the test set accuracy of the improved neural network based on the bit domain feature extraction in each round of iteration. The improved neural network based on the bit domain feature extraction can be basically consistent with the accuracy change curve of the basic neural network on the premise of relieving the bandwidth pressure and improving the system instantaneity. The highest test set accuracy obtained by the improved neural network based on the bit domain feature extraction in the 135 th iteration is 83.5%, which is very close to the highest test set accuracy (85.0%) of the basic neural network. This means that the improved neural network based on bit domain feature extraction ensures the accuracy of target recognition inference while reducing the overall energy consumption and processing time of the system.
The foregoing describes in detail preferred embodiments of the present invention. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the invention by one of ordinary skill in the art without undue burden. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by the person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.
Claims (1)
1. The target distributed fusion identification method based on bit domain feature extraction is used in an edge side communication network topology structure, wherein the edge side communication network topology structure comprises a plurality of edge devices and an edge domain fusion center, and is characterized by comprising the following steps:
a signal acquisition step: obtaining a perception signal from the plurality of edge devices;
a neural network deployment step: the method comprises the steps of obtaining a pre-constructed and trained deep neural network, splitting the deep neural network according to a preset deployment rule, and deploying the deep neural network into the various edge devices and the edge domain fusion center to obtain a distributed neural network architecture;
a signal compression unit is arranged in the deep neural network deployed in the edge equipment, and the signal compression unit is used for extracting bit domain features of an output signal of the deep neural network deployed in the edge equipment to generate bit domain feature vectors;
the edge domain fusion center carries out target recognition according to the bit domain feature vector;
a target identification step: loading the sensing signals into the distributed neural network architecture to obtain a target recognition result;
the signal compression unit comprises a first signal compression layer and a second signal compression layer;
the first signal compression layer adopts a first activation function to map real values of each dimension of an output signal of the deep neural network deployed in the edge equipment to between 0 and 1;
the second signal compression layer is connected with the first signal compression layer, and maps real values of each dimension between 0 and 1 of the first signal compression layer to 0 and 1 by adopting a second activation function to generate a bit domain feature vector;
in the neural network deployment step, the deployment rule is constructed according to the computing capacity and the communication capacity of the edge equipment;
the first activation function is a sigmoid activation function;
the expression of the second activation function is:
wherein y is the output of the second activation function, x is the input of the second activation function, and ε is the adjustment threshold;
the method further comprises the steps of:
calculating average performance: repeatedly executing the target identification step until reaching a preset first time, and acquiring the average performance of the distributed neural network architecture;
and a neural network adjustment step: adjusting the layer number and parameters of the distributed neural network architecture, repeatedly executing the signal acquisition step, the neural network deployment step, the target identification step and the average performance calculation step until reaching a preset second time, taking the distributed neural network architecture with the optimal average performance as an optimal distributed neural network architecture, and adopting the optimal distributed neural network architecture to perform target identification;
the value range of the first times is within the range of 3 to 20;
the value range of the second times is within the range of 50 to 150;
the calculated indicators of average performance include recognition speed and power consumption.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010945655.1A CN112115830B (en) | 2020-09-10 | 2020-09-10 | Target distributed fusion recognition method based on bit domain feature extraction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010945655.1A CN112115830B (en) | 2020-09-10 | 2020-09-10 | Target distributed fusion recognition method based on bit domain feature extraction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112115830A CN112115830A (en) | 2020-12-22 |
| CN112115830B true CN112115830B (en) | 2023-06-02 |
Family
ID=73803665
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010945655.1A Active CN112115830B (en) | 2020-09-10 | 2020-09-10 | Target distributed fusion recognition method based on bit domain feature extraction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112115830B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112818788B (en) * | 2021-01-25 | 2022-05-03 | 电子科技大学 | Distributed convolutional neural network hierarchical matching method based on unmanned aerial vehicle cluster |
| CN114460566B (en) * | 2022-01-04 | 2024-04-30 | 中国科学院声学研究所东海研究站 | Underwater sound target real-time identification method, system and device |
| CN114785792B (en) * | 2022-06-17 | 2022-09-16 | 东云睿连(武汉)计算技术有限公司 | Cloud-edge collaborative video double-path analysis device and method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109525292A (en) * | 2018-12-24 | 2019-03-26 | 东南大学 | A kind of channel information compression feedback method using bit-level optimization network |
| CN111008701A (en) * | 2019-12-03 | 2020-04-14 | 杭州嘉楠耘智信息科技有限公司 | Data quantization method and device based on neural network and computer readable storage medium |
| CN111222532A (en) * | 2019-10-23 | 2020-06-02 | 西安交通大学 | Edge cloud collaborative deep learning model training method with classification precision maintenance and bandwidth protection |
| CN111401412A (en) * | 2020-02-29 | 2020-07-10 | 同济大学 | Distributed soft clustering method based on average consensus algorithm in Internet of things environment |
| CN111445026A (en) * | 2020-03-16 | 2020-07-24 | 东南大学 | Deep neural network multi-path reasoning acceleration method for edge intelligent application |
-
2020
- 2020-09-10 CN CN202010945655.1A patent/CN112115830B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109525292A (en) * | 2018-12-24 | 2019-03-26 | 东南大学 | A kind of channel information compression feedback method using bit-level optimization network |
| CN111222532A (en) * | 2019-10-23 | 2020-06-02 | 西安交通大学 | Edge cloud collaborative deep learning model training method with classification precision maintenance and bandwidth protection |
| CN111008701A (en) * | 2019-12-03 | 2020-04-14 | 杭州嘉楠耘智信息科技有限公司 | Data quantization method and device based on neural network and computer readable storage medium |
| CN111401412A (en) * | 2020-02-29 | 2020-07-10 | 同济大学 | Distributed soft clustering method based on average consensus algorithm in Internet of things environment |
| CN111445026A (en) * | 2020-03-16 | 2020-07-24 | 东南大学 | Deep neural network multi-path reasoning acceleration method for edge intelligent application |
Non-Patent Citations (1)
| Title |
|---|
| 面向边缘计算的分布式深度神经网络研究;邹颖;《中国优秀硕士学位论文全文数据库 信息科技辑》;20200315;第2、13-42页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112115830A (en) | 2020-12-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112115830B (en) | Target distributed fusion recognition method based on bit domain feature extraction | |
| CN103347268B (en) | Based on the self-adapting compressing reconstructing method of energy efficiency observation in cognitive sensor network | |
| CN111713035B (en) | MIMO multi-antenna signal transmission and detection technology based on artificial intelligence | |
| Teganya et al. | Data-driven spectrum cartography via deep completion autoencoders | |
| CN114764577A (en) | Lightweight modulation recognition model based on deep neural network and method thereof | |
| Wu et al. | A holistic approach to reconstruct data in ocean sensor network using compression sensing | |
| Qiu et al. | DeepSIG: A Hybrid Heterogeneous Deep Learning Framework for Radio Signal Classification | |
| WO2023231635A1 (en) | Model transmission method and apparatus | |
| CN114640669A (en) | Edge calculation method and device | |
| Sun et al. | Semantic-driven computation offloading and resource allocation for UAV-assisted monitoring system in vehicular networks | |
| Sirojan et al. | Enabling deep learning on embedded systems for iot sensor data analytics: Opportunities and challenges | |
| Alam et al. | Error-control truncated SVD technique for in-network data compression in wireless sensor networks | |
| Agostini et al. | Federated learning for multipoint channel charting | |
| An et al. | Channel estimation for one-bit massive MIMO based on improved cGAN | |
| Gao et al. | Secure data fusion in wireless multimedia sensor networks via compressed sensing | |
| CN111914923B (en) | Target distributed identification method based on clustering feature extraction | |
| Chen et al. | Deep learning based antenna muting and beamforming optimization in distributed massive MIMO systems | |
| Tao et al. | Recognition of Electromagnetic Signals Based on the Spiking Convolutional Neural Network | |
| Rao et al. | Sparse Federated Training of Object Detection in the Internet of Vehicles | |
| US20230070003A1 (en) | Determining locations of objects in communication systems | |
| Zhu et al. | Second-Order Statistical Modeling Method of Depth Features in Image Classification | |
| Guan et al. | Research on Intelligent Transmission of Space-Based Information | |
| Zhu et al. | A Semantic Communication Based Wireless Image Transmission for Internet of Things Devices | |
| Brito et al. | Compression of Activation Signals from Split Deep Neural Network | |
| Qian et al. | Deep Image Semantic Communication Model for Artificial Intelligent Internet of Things |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |